GPT-Rosalind earns a high biology rating

OpenAI’s GPT-Rosalind-5.5 scored 63.2% on its Labwork Bench versus 55.8% for GPT-5.5, according to the model’s system card. The vendor-run test measures whether a model can link perturbations to outcomes in real wet-lab protocols; the same card classifies Rosalind’s biology and chemistry capability as High, below Critical, and restricts access to approved organizations with governance controls. For teams following the archive’s earlier account of Rosalind as a restricted life-sciences reasoning layer, file the 7.4-point lift as a benchmark signal, not proof of better drug programs, until prospective lab results arrive.